Computational Analysis and Deep Learning for Medical Care. Группа авторов

Чтение книги онлайн.

СКАЧАТЬ 64 64 - 14 × 14 × 512 437K 88M Inception (4c) 14 × 14 - - - - 2 128 128 256 24 64 64 - 14 × 14 × 512 463K 100M Inception (4d) 14 × 14 - - - - 2 112 144 288 32 64 64 - 14 × 14 × 528 580K 119M Inception (4e) 14 × 14 - - - - 2 256 160 320 32 128 128 - 14 × 14 × 832 840K 170M Max pool 14 × 14 - 3 × 3 - 2 0 0 7 × 7 × 832 Inception (5a) 7 × 7 - - - - 2 256 160 320 32 128 128 - 7 × 7 × 832 1,072K 54M Inception (5b) 7 × 7 - - - - 2 384 192 384 48 128 128 - 7 × 7 × 1,024 1,388K 71M Avg pool 7 × 7 - 7 × 7 - - 0 0 1 × 1 × 1,024 Dropout (40 %) - - - 1,024 - 0 - 1 × 1 × 1,024 Linear - - - 1,000 - 1 - 1 × 1 × 1,000 1,000K 1M Softmax - - - 1,000 - 0 - 1 × 1 × 1,000

The architecture is a shortcut connection of VGGNet (consists of 3 × 3 filters) that is inserted to form a residual network as shown in figure. Figure 1.7(b) shows 34-layer network converted into the residual network and has lesser training error as compared to the 18-layer residual network. As in GoogLeNet, it utilizes a series of a global average pooling layer and the classification layer. ResNets were capable of learning a network with a maximum depth of 152. Compared to the GoogLeNet and VGGNet, accuracy is better and computationally efficient than VGGNet. ResNet-152 achieves 95.51 top-5 accuracies. Figure 1.7(a) shows a residual block, Figure 1.7(b) shows the architecture of ResNet and Table 1.7 shows the parameters of ResNet.

      1.2.7 СКАЧАТЬ